Natural genetic variation as a tool for discovery in Caenorhabditis nematodes

Abstract

Over the last 20 years, studies of Caenorhabditis elegans natural diversity have demonstrated the power of quantitative genetic approaches to reveal the evolutionary, ecological, and genetic factors that shape traits. These studies complement the use of the laboratory-adapted strain N2 and enable additional discoveries not possible using only one genetic background. In this chapter, we describe how to perform quantitative genetic studies in Caenorhabditis, with an emphasis on C. elegans. These approaches use correlations between genotype and phenotype across populations of genetically diverse individuals to discover the genetic causes of phenotypic variation. We present methods that use linkage, near-isogenic lines, association, and bulk-segregant mapping, and we describe the advantages and disadvantages of each approach. The power of C. elegans quantitative genetic mapping is best shown in the ability to connect phenotypic differences to specific genes and variants. We will present methods to narrow genomic regions to candidate genes and then tests to identify the gene or variant involved in a quantitative trait. The same features that make C. elegans a preeminent experimental model animal contribute to its exceptional value as a tool to understand natural phenotypic variation.

Keywords: Caenorhabditis; QTL mapping; WormBook; genetic variation; quantitative genetics; recombinant inbred lines.

Figure 1

QTL-mapping panels. Different experimental designs make different compromises along many axes, including detection power, mapping resolution, and genetic diversity. Laboratory crosses of a pair of strains provide a straightforward route to mapping by linkage in RILs, represented here as a single chromosome from each of six RILs from a cross of two strains, one with an orange genome and the other blue. Because of the low recombination rate in C. elegans, RIL chromosomes have an average of one crossover each. By adding generations of intercrossing prior to inbreeding, RIAILs increase the number of breakpoints, increasing mapping resolution. Association mapping uses historical recombination to randomize alleles. Wild isolates carry a mixture of common alleles that arose in ancient ancestors (black symbols) and more recent alleles that are unique to each strain (red stars). The pattern of association among shared variants (LD) is governed by the population history of recombination, and some variants may be perfectly correlated (e.g., black star and black hexagon). Multiparent panels use laboratory crosses to shuffle wild isolate genomes even more, reducing LD and increasing the frequency of rare alleles; now even singletons (red stars) are visible to QTL mapping. QTL can also be discovered by comparing strains that differ only in a small interval—near-isogenic lines. Finally, bulk segregant and related evolve-and-resequence methods do away with the construction of inbred lines. They detect QTL as differences in allele frequencies between pools of individuals selected to differ in phenotype. In the figure, the allele frequencies differ between the high-phenotype pool and the low-phenotype pool in the highlighted interval.

Figure 2

Epistasis for free. A hypothetical example of a trait with pure redundancy between two underlying loci that differ between the strains N2 and AB1. Of 100 simulated RILs, only those carrying AB1 alleles at both loci show a phenotypic effect of the loci (top panel). But considering the loci one at a time (middle and bottom panels), each locus has a clear marginal effect and is easily detectable by QTL mapping methods that do not explicitly consider epistatic interactions. Note that the power to detect these loci is strongly influenced by the frequency of the rare phenotypic class (here, ¼, but often much lower in GWA mapping designs). Methods tailored to discover epistatic interactions often take advantage of the difference in phenotypic variance between single-locus phenotypic distributions to identify candidate interactors (Struchalin et al. 2010; Rönnegård and Valdar 2012).

Figure 3

RIL construction. Conventional RILs are generated by crossing two inbred lines (P₀), each here represented by a single pair of homologous chromosomes. The cross yields F₁s that are heterozygous at every locus that differs between the strains. Self-fertilization of F₁ hermaphrodites yields genetically heterogeneous F₂s. Because of complete crossover interference in C. elegans, each F₂ carries on average one recombinant chromosome with a single crossover. Each F₂ is independently inbred by selfing until the F₁₀ generation, at which point the genome is expected to be completely homozygous, with an average of one crossover per chromosome. This collection of homozygous strains is a panel of RILs. Note that this panel will all inherit the mitochondrial genome of the P₀ hermaphrodite. To balance the panel with respect to mitochondrial genotype, an equal number of RILs should be constructed from the reciprocal cross.

Figure 4

Linkage mapping power and example. (A) The statistical power of different RIAIL panel sizes is plotted by the QTL effect (percent of V_P that is V_QTL). Larger RIAIL panels can identify QTL that in total explain more of the total phenotypic variance. These data were generated using the QTLDesign package in R (Sen et al. 2007). (B) An example linkage mapping plot for response to zinc (Evans et al. 2020) is shown. Genomic position (x-axis) is plotted against the logarithm of the odds (LOD) score (y-axis) for 13,003 genomic markers. Each significant QTL is indicated by a red triangle at the peak marker, and a blue rectangle shows the 95% confidence interval around the peak marker.

Figure 5

Genome-wide association mapping power and example. (A) Statistical power for the 20200815 CeNDR release strain set is plotted by the QTL effect (Percent of phenotypic variance explained by the QTL). (B) A Manhattan plot for single-marker-based GWA mapping of the ascr#5-induced dauer formation trait (Lee et al. 2019) is shown. Each dot represents a single-nucleotide variant (SNV) that is present in at least 5% of 157 wild strains. The genomic position in Mb, separated by chromosome, is plotted on the x-axis, and the statistical significance of the correlation between genotype and phenotype is plotted on the y-axis. Two significance thresholds are shown. The dashed horizontal line denotes the Bonferroni-corrected P-value threshold using all markers, and the solid horizontal line denotes the Eigen-corrected P-value threshold using independent markers correcting for LD (genome-wide eigen-decomposition significance threshold). SNVs are colored red if they pass either threshold.

Figure 6

Genetic causality experiments in C. elegans. Step 1 is to identify QTL using statistical mapping approaches. A GWA mapping is shown here. Step 2a uses NILs to narrow the QTL confidence interval found in Step 1, shown as dotted red vertical lines. In parallel or iteratively, QTL fine-mapping approaches like Step 2b can also narrow a QTL to candidate genes. Once candidate genes are identified, genome-editing using CRISPR-Cas9, as shown in Step 3a, can replace the candidate gene allele from Strain A with the Strain B allele and vice versa. Sometimes candidate genes have many different variants or no single variant can be tested as in Step 3a. In these cases, a loss-of-function allele like a deletion shown in Step 3b can be created in both Strain A and Strain B genetic backgrounds using CRISPR-Cas9 genome editing. Then, these new deletion strains can be crossed to the reciprocal parent strains and compared to the parent strains and heterozygotes in a quantitative complementation experiment. In the example shown here, Strain A (orange) has a low trait value and Strain B (blue) has a high trait value. The Strain B allele confers a dominant phenotype as seen in the heterozygote. Deletion of the Strain A allele has no effect on the phenotype of the heterozygote but loss of the Strain B allele fails to complement the Strain A allele. These data indicate that the gene deleted is the same gene that confers a trait difference between these two strains.